Home › Topics › The Fundamentals That Outlive Every Tool › IPv6 Readiness

IPv6 Readiness: Monitoring a Network That Is Already Dual-Stack


You may believe your network does not use IPv6. I believed that about a network once, right up until I ran one command on a laptop. I found it holding three IPv6 addresses and a default route to a device I had never heard of. Its neighbour table listed half the floor. Nobody had configured any of it. That is the thing about IPv6: it does not wait to be invited.


This matters because your monitoring system almost certainly checks the IPv4 side of everything and nothing else. Meanwhile every modern laptop, server, and phone on the network carries an IPv6 address and prefers it when given the chance. Each device will find a service broken over a path your monitor has never looked at. The check stays green. The users do not.


By the end of this guide you will be able to explain IPv6 in plain words. You will be able to say why it affects your monitoring even if you never turned it on, and name the blind spots it creates. You will be able to run a handful of commands to see what is already on a host. And you will be able to put your monitoring system through a readiness checklist you can copy. None of it requires you to deploy IPv6. It requires you to notice that it was deployed for you.


In this guide

  • IPv6 in plain words
  • Why it matters even if you "don't use IPv6"
  • The blind spots it creates
  • Seeing what is already on a host
  • What to ask of the monitoring system
  • The IPv6 readiness checklist
  • The intranet that was down for every laptop
  • The checklist
  • Where to go next

IPv6 in plain words


IPv6 is the newer of the two ways of numbering things on a network. The older one, IPv4, uses addresses 32 bits long, written as four numbers separated by dots, like 192.0.2.10. And there were never enough of them. IPv6 uses addresses 128 bits long, written as eight groups of hexadecimal digits separated by colons. There are so many that nobody has to be careful with them. A double colon stands in for a run of zeros, so 2001:db8:0:0:0:0:0:10 is written 2001:db8::10. Both of those example ranges are reserved for documentation, which is why you will see them in every guide and never on a real network.


There are several kinds of IPv6 address, and one of them is the reason this guide exists. A link-local address begins with fe80:. Every network interface makes one for itself the moment it comes up, with no server and no configuration. It works only on the segment the interface is on and is never routed anywhere. A global address begins with a 2 or a 3 and can be routed across the internet. A unique local address begins with fd and is the IPv6 equivalent of a private range. The loopback address is simply ::1. A link-local address written with a percent sign after it, such as fe80::1%12, names the interface it belongs to, because every interface has its own link-local range.


Global and unique local addresses come from one of two places. A router on the segment sends a router advertisement. This is a small message that says "I am a router, and this is the address range here". Each host builds its own address from that range without asking anyone, a process called stateless address autoconfiguration. Or a DHCPv6 server hands addresses out, as DHCP does for IPv4. Hosts find each other's hardware addresses with neighbour discovery, which replaces ARP. All of these messages travel over ICMPv6, the IPv6 version of the protocol that carries ping. That is why a firewall that blocks ICMPv6 wholesale does not harden IPv6 so much as break it.


Dual-stack means a host runs both protocols at once, with an IPv4 address and one or more IPv6 addresses on the same interface. The host chooses which to use for each connection. The choice is driven by DNS. A name with an A record resolves to an IPv4 address. A name with an AAAA record resolves to an IPv6 address; a name can have both. When it does, most operating systems prefer IPv6. Many web browsers soften this by trying both and using whichever answers first, which hides a broken IPv6 path from the browser's user. Most other programs simply try IPv6 first and wait for it to fail, which hides nothing.


Why it matters even if you "don't use IPv6"


Three mechanisms put IPv6 on your network without a project, a budget, or a decision, and each creates a path your monitoring is not watching.


Every modern device already has an address

Every operating system you are likely to meet enables IPv6 by default and gives each interface a link-local address as soon as it comes up. Devices on the same segment then find each other over IPv6 and talk without being asked. That includes name resolution by multicast, printer and file-share discovery, and the rest. Two servers on the same segment can hold an entire conversation over link-local addresses that appear in no inventory and no map. Nobody assigned them. You do not use IPv6 in the same sense that you do not use the stairs. They are still there, and people are still on them.


Dual-stack hosts prefer IPv6, and the check does not

A service acquires an AAAA record more easily than you would think. A server administrator leaves IPv6 enabled on a new server. The server registers its own addresses in DNS as Windows servers do by default. From that moment every dual-stack client tries the IPv6 address first. Suppose the IPv6 path is broken — a firewall with no IPv6 rules, a missing route, or a segment where the advertisement lies. The client waits for a timeout and then may or may not fall back. Your monitoring check, meanwhile, pings the IPv4 address it was given on the day it was created, and it is green. The service is up on the path the monitor uses and down on the path the users use. Ping up versus service up describes a cousin of this problem.


Rogue router advertisements

Any device that sends router advertisements on a segment becomes an IPv6 router for every host on it, because hosts believe advertisements by default. A home router plugged in the wrong way round, a virtual machine with a misconfigured network service, or a phone sharing its connection will do it. Every host on the segment gets a new global address and a default route pointing at the rogue. And the preference rule then sends their traffic to it. The rogue device detection pillar covers finding the culprit. Managed switches usually offer a feature that drops advertisements from ports that should not send them. And it is worth switching on before you need it. (The rogue is almost never malicious. It is almost always someone's desk.)


The blind spots it creates


A monitoring practice built around IPv4 has five specific gaps, each fixed differently.


  • Discovery that only scans IPv4 ranges. Auto-discovery usually works by sweeping an address range, and an IPv6 segment is far too large to sweep. A single standard segment holds more addresses than the entire IPv4 internet. IPv6 discovery has to work from neighbour tables, DNS records, and DHCPv6 leases instead. If your tool only sweeps, it has only ever seen half of every host.
  • Flow records you ignore. Routers and switches export flow records for IPv6 traffic as readily as IPv4, provided they use one of the newer, template-based export formats. The oldest format carries IPv4 only. If your reports filter on IPv4 addresses, the IPv6 traffic is in the database and absent from every graph. Flow monitoring is the pillar for what the records contain.
  • Firewall rules that forgot the other protocol. A rule that permits or blocks a host's IPv4 address says nothing about the same host's IPv6 address. Some firewalls default to blocking all IPv6, which breaks dual-stack services silently. Some default to allowing it, which bypasses every IPv4 rule you ever wrote. Neither shows up in an IPv4 check.
  • DNS records nobody monitors. An AAAA record pointing at an address no check has ever tested is a promise the monitor cannot see. DNS troubleshooting covers the resolution side; the monitoring side is to know which names have AAAA records and test each one.
  • Inventories and maps with one address column. A device record with room for one address records the IPv4 one. And the IPv6 address the same device uses for half its conversations is nowhere. Network blind spots catalogues the general problem; this is its most common specific case.

Seeing what is already on a host


Before you change anything in the monitoring system, look at one host, and the laptop you are reading this on is a fine start. The commands below show every IPv6 address it holds and any default route it has learned from an advertisement. They show the neighbours it has talked to and whether a name you care about has an IPv6 record at all. On Windows:


ipconfig /all
    (every address on every interface, both families)

netsh interface ipv6 show addresses
    (IPv6 only, with the type: link-local, global, temporary)

netsh interface ipv6 show route
    (a ::/0 route is a default route learned from an advertisement)

netsh interface ipv6 show neighbors
    (the IPv6 neighbour table: who this host has talked to)

nslookup -type=AAAA intranet.example.com
    (does the name have an IPv6 record?)

ping -6 intranet.example.com
    (force the test over IPv6)

On a Unix-like system the same questions look like this:


ip -6 addr
    (every IPv6 address, with its scope: link, global, host)

ip -6 route
    (the default route, if one has been learned)

ip -6 neigh
    (the neighbour table)

dig AAAA intranet.example.com +short
    (the IPv6 record, if any; empty means IPv4 only)

ping -6 intranet.example.com
    (force IPv6; older systems spell it ping6)

Read the route output most carefully. A default route pointing at a link-local address you do not recognise is an advertisement from a device you did not intend to be a router. That device is worth finding today. Read the neighbour table second. If it lists a dozen addresses on a segment you believed was IPv4-only, the segment is not IPv4-only, and it never was. And if the name has an AAAA record and the IPv6 ping fails while the IPv4 one succeeds, you have found the exact situation this guide is about. You have found it with one command the monitoring system has never run.


What to ask of the monitoring system


Monitoring tools handle IPv6 unevenly. Some poll over it, discover it, and check services over both families as naturally as IPv4. Some support it in the poller but not in discovery. Some record IPv6 addresses but never test them. There is no way to know which yours is except by asking. For every question below the honest instruction is the same. Check the product documentation, then test it against a real host.


Can it poll over IPv6?

Polling over IPv6 needs two things. The device's SNMP agent, WMI service, or SSH daemon must listen on an IPv6 address. The poller must be able to reach it over IPv6. Many agents listen on IPv4 only unless told otherwise, so a "yes" from the tool is only half the answer. Test with a device you know listens on both, added to the tool under its IPv6 address literal.


Can discovery find IPv6 addresses?

Ask whether discovery reads neighbour tables and DNS, or only sweeps ranges. A tool that sweeps can still be useful if it discovers a device by IPv4 and then reads the device's IPv6 addresses from the device itself over SNMP. The test is whether a device record ends up holding both addresses after discovery runs.


Do service checks test both address families?

A web check configured against a hostname resolves the name and connects over whichever family the monitoring host prefers. That tests one path and leaves the other untested. One remedy is two checks per dual-stack service — one against the IPv4 address literal and one against the IPv6 literal. The other is a tool that runs both and reports each separately. Either way, the address family belongs in the check's name.


Does the address inventory record both?

The device record, and the address plan behind it, need a place for every address a device holds, in both families. IPAM basics explains address management from the start. The guide to documenting address space shows the record format. The addition for IPv6 is a second column, plus a note on each segment record of which prefix its router advertises. Read subnet planning before you assign anything, because IPv6 segments follow a different rule than IPv4 ones.


Do the maps show it?

A map that draws a link between two devices because they exchanged IPv4 traffic will not draw the link they use for IPv6. A map with a gap in it has taught the room something false. Ask whether the map's link discovery and its labels include the IPv6 side. If they do not, add the address to the label by hand for the devices that matter. That way, the map at least admits the second address exists.


Tip: the monitoring server is itself a dual-stack host, and it will prefer IPv6 for any name with an AAAA record. That includes the devices it polls and the mail server it sends alerts through. If its own IPv6 path is broken it can lose devices or alerts for a reason that appears in none of its logs. Check its addresses and routes before anyone else's.


The IPv6 readiness checklist


Copy the table and work down it with the product documentation in one window and a real dual-stack host in another. Fill in the last column with an honest answer and where the evidence lives. A row you cannot answer is currently "no".


QuestionHow to test itAnswer and where the evidence is
Does the monitoring server hold a global IPv6 address and a default IPv6 route?Run the address and route commands on the server 
Can the poller reach each type of device over its IPv6 address?Add one device of each type by IPv6 literal; confirm a successful poll 
Does discovery record IPv6 addresses on device records?Run discovery; inspect one dual-stack device record 
Which service names have AAAA records?Query each service name for an AAAA record; keep the list 
Does every service on that list have a check against its IPv6 address and one against its IPv4 address?Compare the AAAA list with the check list, both families 
Does the address inventory have a column for IPv6?Open the inventory; look for the column 
Does each segment record name its advertised IPv6 prefix, if any?Open one segment record; compare with the route command on a host there 
Do flow reports include IPv6 traffic?Filter a flow report by an IPv6 address you know is active 
Do firewall rules exist for the IPv6 side of every published service?Review the rule set for each service's IPv6 address 
Do maps show the IPv6 address on dual-stack device labels?Open the map; find one dual-stack device 

Ten rows is a morning's work if the documentation is good and a week's if it is not. The difference between those two is itself something you have learned about the tool.


The intranet that was down for every laptop


Harbourview Clinic replaced its intranet server, and the new one was built with IPv6 left enabled, as the operating system prefers. The server registered its own addresses in DNS. So the intranet's name gained an AAAA record on the first afternoon. It pointed at a server segment whose firewall had never been given a single IPv6 rule. Every dual-stack laptop in the building tried IPv6 first and waited for the connection to time out. It either gave up or fell back to IPv4 half a minute later, depending on which program was asking. The monitoring check fetched the intranet page over the IPv4 address it had been given, every minute, and reported a month of perfect availability. The helpdesk queue for that month said "intranet slow, sometimes down" forty times. The availability report said one hundred per cent, which was true of the path nobody used. A technician found it in ten seconds with one AAAA lookup, and the fix was one firewall rule and one extra check.


Every part of that story is ordinary, which is the point. Nobody deployed IPv6 and nobody decided anything. Four defaults and a firewall with no opinion about the second protocol combined into an outage the monitoring system was structurally unable to see. The monitoring system had only ever been told about one of the two paths.


The checklist


Work through this once now and again whenever a server, segment, or service is added.


  • You have run the address, route, and neighbour commands on one host per segment and kept the output.
  • Every default IPv6 route you found points at a router you intended to be one.
  • You have a list of every service name that has an AAAA record.
  • Every service on that list has a check against its IPv6 address and one against its IPv4 address.
  • The monitoring server's own IPv6 addresses and routes are known and recorded.
  • The address inventory has an IPv6 column, and dual-stack devices have it filled in.
  • Firewall rules for published services have been reviewed for the IPv6 side.
  • Advertisement filtering is enabled on access ports where the switches support it.
  • The readiness table is filled in, with an answer on every row.
  • The new-server build procedure includes "check for an AAAA record and add the second check".

Where to go next


IPv6 readiness for monitoring is not a migration and not a project. It is the admission that every host already has a second set of addresses and that clients prefer them. A check which tests one family is telling you about half a network. Run the commands on one host today, list the names with AAAA records, add the second check for each, and fill in the table. After that, the monitor sees what the users see, which was always the point.


Testing what users feel rather than what the check was told is principle one in the ten principles. "Is it dual-stack, and do we check both?" is one of the questions in When everything changes, because a merger or a new job is exactly when a second address family arrives unannounced. Both are part of The Fundamentals That Outlive Every Tool.


Put this into practice with SysPeep Monitor

SysPeep Monitor polls devices over SNMP, WMI, and SSH from a central Windows service. It auto-discovers devices and interfaces, and shows them on maps and dashboards in one web console. For the IPv6 specifics in this guide, check the product documentation. Those specifics include polling over IPv6, discovering IPv6 addresses, and running service checks against both address families. Use the free trial to test the answers against one real dual-stack host before you rely on them.

Download Free TrialLearn About SysPeep